DIME

Direct and Inverse Modelling in
End-to-End Environmental Estimation

The end-to-end approach aims to track uncertainty, both from model inadequacy and from the unknown initial state of the atmosphere, all the way through the modelling process, to yield estimates of the uncertainty In quantities of industrial interest. We will contrast, and ideally meld, both direct (forward or physical simulation) models and inverse (statistical or empirical) models in a variety of contexts. London Electricity, one of our industrial collaborators, is most interested in the time-scales of days; here we will examine methods for forming hybrid ensembles from ensembles over initial condition, model, and initialisation time. Hybrid ensembles dress the individual trajectories of a Monte Carlo ensemble of trajectories with appropriate historical error statistics, providing a user specific approach to downscaling. More generally, the use of direct and inverse models will be contrasted in downscaling applications associated with wind energy. New inverse models for seasonal models will be constructed and compared with direct seasonal forecasts from the DEMETER project; the comparison will be based on industrial utility in quantifying weather risk. This is of direct industrial interest to Risk Management Solutions, our other industrial collaborator.

Final report (pdf)

Share:Facebook|Twitter|LinkedIn|